Selecting an Instrument
In general, you can choose between two types of PROMIS® instruments
for your study: computerized adaptive tests (CATs) or static short forms. Your choice
should be driven by your relative interests in precision, brevity, item content,
and flexibility/portability. No matter which instrument and PROMIS items you choose,
the end result will be a common (Theta) metric which has been converted to a T–distribution
based on the United States general population.
Precision can be increased by adding questions from the same item bank. However,
there are two important things to keep in mind. First, incremental gains in precision
decrease as the overall number of questions increases. In other words, while a 6–item
scale is much more precise than a 1–item scale, the gain in precision realized by
adding 5 more items (to create an 11–item scale) is not nearly as much as the gain
in precision realized when the previous six items were added. Second, not all items
in an item bank are equally informative, so item selection matters. This is the
reason that CAT will almost always be more precise than a fixed short form of the
same length. PROMIS banks using CAT can achieve precision that meets standards for
individual level assessment in fewer than 6 items on average, while static short
forms require many more items to obtain the same level of precision. In cases
where precision is your main goal (for example, when tracking an individual person
over time to detect reliable change), a CAT or a static short form of 10 or more
items should be preferable.
While both CAT and static short forms can be brief, CAT will out–perform a static
short form of the same length. So, if brevity and precision are desired, CAT is
the better choice of instrument. However, there may be certain settings where brevity
is desired, but CAT is neither possible nor desired. Additionally, some applications
(e.g., large sample studies seeking population estimates; large sample clinical
trials that plan group comparisons) do not require the precision offered by CAT
or lengthy short forms. In those cases, carefully selecting a small number of items
per bank (as few as one per bank) will produce T–scores that can be referenced to
the general population sample. Although these short forms may result in individual
score estimates of low confidence, large group averages are reliable. Use the item
statistics in Assessment Center to
create customized short forms from PROMIS item banks. We have created sample short
forms for each item bank, ranging from 6 to10 items per form. These forms contain
some of the more informative items that cover the measurement range. Each form is
suited for individual assessment in the middle range of the trait being measured.
Extremes on the measurement continuum (e.g., very little fatigue or extreme fatigue)
provide estimates of lower reliability. Briefer short forms can be custom made.
Sometimes, researchers may prefer to determine which questions in a bank are administered
based on the clinical relevance of a desired subset of items, or the lack of relevance
of a subset of questions to a given target research population. Similarly, some
researchers may wish to ensure that the same items get administered at certain time
points in a longitudinal design. In this case, static short forms should be the
preferred instrument since only one item in the bank administered by the CAT engine–the
first one–would be guaranteed to be repeated at each time point. CAT options do
exist, however, to "balance" content of items administered from a given
bank, which is important to remember before opting out of the CAT approach.
CAT requires access to a computer (for either Web based or standalone administration).
Although options exist for branched assessment that approximates CAT on paper, these
options are not currently available for PROMIS item banks. Thus, if you do not have
the capability to electronically enter participant responses in real time, you should
select paper (or telephone) administration of static short forms.